The handle http://hdl.handle.net/1887/80957 holds various files of this Leiden University dissertation.
Author: Flöthe, L.
Title: Speaking for the People? Analysing the extent to which interest groups represent the opinion of the citizens and under which conditions they are more likely to do so
Chapter 3: Representation through
Information
The article for this chapter has been published in the Journal of European Public Policy as:
88
Representation through information?
When and why interest groups inform policymakers about public
preferences
Abstract
89
Acknowledgments
90
3.1 Introduction
91
gauging fine-grained arguments to uncover some of the underlying dynamics of information provision.
The results show that, overall, interest groups provide information on public preferences. However, citizen groups do so more frequently than professional groups, business groups and experts, suggesting that they have a greater potential to act as transmission belts. Moreover, predominantly opponents of policy change transmit public preferences in order to protect these interests if they are at risk. Furthermore, actors who share the same opinion as a large part of the public inform policymakers more about these preferences. The contribution of this paper is twofold. First, it conceptualises interest groups as transmission belts and provides an empirical test of the assumption that interest groups inform policymakers about what the public wants which is a necessary condition for them to act as transmission belts. Second, it highlights conditions under which actors are more likely to do so which contributes to our understanding of when interest groups have the potential to help governments to respond to public preferences.
3.2 Interest Groups as Transmission Belts
92
necessarily mean information about members of a group, yet refers to certain subparts of society such as ‘families’ or ‘the poor’.
The literature on informational lobbying has referred to such information as political information, which includes information regarding support or opposition of a specific constituency or the public at large (see for example De Bruycker 2016; Nownes and Newmark 2016). Importantly, however, Nownes finds that advocates do not necessarily make arguments about the public as whole, but rather about certain parts of society (2006: 66). To allow for a systematic analysis of how interest groups can act as transmission belts, the paper defines such information as information on public preferences, which refers both to information on preferences of the public at large but also preferences of specific constituencies and certain segments of the society (cf. Burstein 2014).
In order to understand how groups act as transmission belts the paper follows Saward who defines representation as a dynamic process in which multiple actors articulate representative claims to an audience to ‘represent or to know what represents the interests of someone or something’ (2006: 305). Saward criticises Pitkin (1967), who acknowledges that representative institutions provide information about the people, but takes such information as given and neglects the process of providing such information. Saward shifts the focus to the act of making present and the actor making such claims. Even though Saward’s conceptualisation is not without problems either, the focus on claims allows for analysing representation through non-elected representatives such as interest groups (for a discussion, see De Wilde 2013).
A representative claim can be expressed in a number of ways but may refer for example to the needs/desires/preferences of a person or a group of people. Representation through interest groups, then, can be thought of as an act where advocates mobilise on a specific issue (e.g. reforming child support) to actively promote a position (e.g. no cuts) in the interest of a group of people (e.g. families with children) by informing policymakers about the interests of the group of people (cf. Severs 2012). So for representation to occur and for a group to act as a transmission belt, advocates may either signal support or opposition of the public at large or, importantly, of specific constituencies. Such a conceptualisation considers the two underlying mechanisms of how the transmission belt works. A first assumption therefore is:
93
While this does not allow for drawing inferences about whether interest groups are effective in transmitting preferences, it sheds light on a necessary (but insufficient) condition for acting as a transmission belt, i.e., whether (and under which conditions) they provide such information in the first place. Given the focus on the actors of ‘making present’, the paper theorises how variation in the actor’s characteristics affects information provision.
3.2.1 Who informs about Public Preferences?
Although scholars have not found differences across actor types with regard to information provision (De Bruycker 2016; Nownes and Newmark 2016), there are several hints in the literature why we could expect groups to differ in their motivations for transmitting information about public preferences.
First, information provision is determined by the type of constituency. While some groups have a clearly defined constituency, others represent a broad public interest (Olson 1965). For example, some groups aim at promoting broader interests (improved air quality) which are not tied to a specific constituency (such as doctors) or restricted to benefit members only (Binderkrantz et al. 2015). Instead, the benefits are collectively available. Public interest groups typically defend diffuse public interests that are not exclusive to their members but the public at large (Dür and Mateo 2013). Even identity groups (e.g. patient groups), who have a slightly more specific constituency promote interests that also non-members could benefit from. Since these groups often rely on (potential) members and supporters for organisational survival, they are under greater pressure to demonstrate that they act in the interest of their constituency (Flöthe and Rasmussen 2019; Klüver 2015: 141), which may also increase the transmission of information about their preferences.
94
business groups, their main motivation is to primarily provide services that mainly their members would benefit from.
Secondly, groups have different capacities and exchange goods to offer when lobbying policymakers (Bouwen 2002; Daugbjerg et al. 2018; Dür and Mateo 2013). Policymakers need technical expertise to increase their output legitimacy, but also information about political support to increase their input legitimacy (Bouwen 2002; Wright 1996). While information about general public opinion may be quite accessible for policymakers, issue-specific information about preferences of different sub-groups is more difficult to access. Policymakers may have preferred options for sources for the different types of information. Citizen groups, since they represent broad interests, are assumed to articulate a diversity of interests and are therefore able to contribute to the input legitimacy of the policymaking process (Kohler-Koch 2010: 106). Moreover, they should validate that their claims reflect the concerns of their constituents (Kohler-Koch 2009: 54) and invest in ‘determining member preferences’ (Schlozman and Tierney 1986: 142), which makes them a credible source of such information and can help to legitimise a policy decision (Michalowitz 2004: 85).
Actors without mass membership such as expert organisations but also firms cannot credibly provide this information to the same extent (cf. Wright 1996: 92). In contrast, business groups, professional groups and also experts are a credible source of expert information (Bouwen 2002) as they are close to the market (Dür and Mateo 2013; Eising 2007), have hands-on experience (Dür and Mateo 2013; Michalowitz 2004), better capacities to understand the technical and scientific context or even such data themselves (Yackee and Yackee 2006). Given that each information type requires different resources, one can expect groups to specialise in their core capacities (Daugbjerg et al. 2018) and those resources they are especially approached for. This does not imply that business groups or professional groups do not provide information about their members, yet given that access to expert information is easier for them (Dür and Mateo 2013) one could expect them to emphasise this type of information more. Similarly, citizen groups might focus on their core capacity, i.e., provide input legitimacy by transmitting information about public preferences.
95
interests. This does not rule out that all actor types transmit information on public preferences but their propensity to do so should vary.
H1: Citizen groups are likely to provide more information on public preferences than professional groups, business groups and expert groups.
3.2.2 Under which Conditions do Actors inform about Public Preferences?
Interest groups may not necessarily transmit information on public preferences with the intention to represent the public’s interest, but to strategically justify their position and pressure policymakers (De Bruycker 2016; Wright 1996). Research on informational lobbying shows that actors lobby differently depending on their position on a policy (Baumgartner et al., 2009; Burstein 2014). Burstein shows that opponents of policy change use arguments that cast doubts regarding the proposed solution and its effectiveness (2014: 148), suggesting that opponents use information negatively to warn for undesired consequences.
This suggests that advocates channel their constituents’ interests especially when their interest is at risk. Kingdon noted that the public sometimes directs governments to do something, yet most of the time constrains the government from doing something (1984). Policymakers rely on interest groups for information to reduce some of the uncertainties they face when deciding on a policy (Wright 1996) and opponents of policy change can use information to highlight the risks of change, something policymakers fear. Hence, opponents of policy change use information on public preferences more frequently to signal negative consequences for (parts of) the public. They may transmit such information to warn of negative consequences to strategically increase fears and uncertainties or to protect their constituents for undesired policy change.
H2: Opponents of policy change transmit information on public preferences more frequently than supporters.
96
difficult for the government to ignore. Knowing how important public support is, interest groups cannot ignore it (Nownes 2006: 101) and may even be tempted to use information about public preferences strategically (Wright 1996). The more people the actor has on its side, the higher the representational value of an actor’s claim as a large part of the public may benefit from or support the new policy. It does not mean that actors who represent minority preferences do not transmit their constituents’ interest when they only have low support for their claim. However, the representational value would be rather low and the electoral consequences for policymakers may be minor. In such a scenario, the emphasis on this information should be limited at best. Likewise the likelihood of transmitting more of the information should increase if the actor enjoys broad public support as it demonstrates broad acceptance for the claim.
H3: An actor with a higher proportion of the public on its side is likely to transmit more information on public preferences than an actor with lower support.
3.3 Research Design
The study relies on observed information transmission in public hearings held by standing committees of the German parliament. The hearings are supposed to generate issue-specific expertise and information on actors’ position and general support on the topic (Burstein 2014: 130; Eising and Spohr 2017: 316). Public hearings in Germany take place after a bill proposal has been assigned to a committee. The proposals have been initiated by the government and opposition parties. Eventually, 59% of the issues in the sample were enacted. One could argue that predominantly advocates are invited to hearings who support a policy proposal to help policymakers legitimise a policy decision.20 Yet, we see that only 36% of the advocates were in favour of policy change. This indicates a higher mobilisation of actors who want to protect the status-quo and speaks to a common pattern found in the literature (cf. Baumgartner et al 2009).
As a working parliament, the German Parliament consists of highly professionalised parliamentarians who are able to exert a considerable degree of influence which makes the committees and their hearings an important venue for advocates (Eising and Spohr 2017: 318-9) and hearings an interesting case to uncover underlying mechanisms of information
20 Also submissions of uninvited actors have been coded. However, they only account for a small share (4%) and
97
provision. Analysing written statements by interest groups is a novel way of studying information provision. Most studies rely on self-reported information transmission through surveys or interviews (for an exception see Burstein 2014). This is likely to uncover only the information types that actors find most important and may overlook types that are less consciously used. Focussing on one venue in a single country allows labour-intensive coding of arguments to get a more accurate picture of information provision (cf. Burstein 2014: 130-59).
3.3.1 Issue Sampling and Data Collection
The sample of issues is based on a dataset developed within the larger GovLis21 project that contains 102 specific policy proposals in Germany. The starting points of data collection are existing nationally representative public opinion polls on specific policy issues that were held between 1998 and 2010. Selected issues had to fall under national jurisdiction (as opposed to the EU or sub-national level) and the opinion poll questions had to ask for a change of the status quo. Issues in the sample concern, for example, the question of raising the tobacco tax (see Appendix A for a list of issues). Polls are likely to be conducted on relatively salient policy issues and a sample based on them does not constitute a completely random sample of policy issues (Burstein 2014). However, citizens should have at least somewhat informed opinions if interest groups are expected to transmit them meaningfully (Gilens 2012: 50-6). Following Gilens (2012), the observation period for each policy issue starts in the year the policy item was asked by the pollster and ends four years later or when policy changed. This study relies on a subsample of issues on which public hearings were held during the observation period and focuses on written evidence. The final sample contains 34 issues on which 42 hearings were held in which actors made 356 statements about the issue. The unit of analysis is an actor in a hearing. Each actor is counted once for testifying at a hearing on the specific issue, however appearances at different hearings on the same issue are counted separately (Burstein 2014: 141; Eising and Spohr 2017). An issue can be discussed in multiple hearings and a hearing can discuss multiple issues, which suggests a cross-classified multilevel structure with actors nested in hearings and issues. However, given that variance at the hearing level is quite low and the data structure overly complex, information provision will be modelled in a two-level structure with actors nested in policy issues. Model fit does
98
not significantly differ irrespective of whether actors are nested within hearings or issues or within each other.
3.3.2 Dependent Variable
Information types in this paper are conceptualised as arguments with which advocates underpin their position. Arguments are stated reasons an actor uses to justify and substantiate its position (Eising et al. 2017: 5). Thus, after identifying an actor’s position on an issue, the different types of arguments used to defend the position were coded by two trained coders. Appendix C1 contains the coding instructions including examples. Arguments are counted separately if a different argument is provided in the next paragraph or if the causal story for why the actor supports (opposes) policy change differs. Two coders independently coded 50 units, which resulted in an acceptable Krippendorff’s alpha of .72 (De Wever et al. 2007).
The dependent variable Information on Public Preferences relies on two proxies which capture the underlying mechanisms of the transmission belt. The first proxy counts how often an actor makes any references about how much public support (opposition) a policy proposal has. The second proxy records how often the actor argues how a policy proposal will affect certain segments of society. This partially follows Burstein’s operationalisation of political information, which includes not only references to broad public support but also how advocates refer to how a policy will affect certain subparts they (claim to) represent. An example would be ‘We oppose the proposal because it will aggravate the situation of the poor’. This measurement allows gauging the observed transmission of more specific information that interest groups provide about constituency preferences (as opposed to general public opinion polls that policymakers can also access via other channels). The count measure moreover captures the extent to which actors reinforce certain arguments. The dependent variable combines these two count measures and ranges from 0 to 11 (see Appendix B1 for an overview of all variables).
3.3.4 Independent Variables
99
category ‘citizen groups’ includes public interest groups and identity & hobby groups such as environmental groups or patients groups. This category includes groups that have a strong incentive to represent their members or a more diffuse interest. The category ‘business groups’ includes firms and business associations, which have a strong advantage over policy expertise and which aim to protect exclusive interests. Professional groups include trade unions and occupational groups. A last category refers to experts, institutional associations and think tanks that are assumed to provide expertise without taking a side for a preferred constituency. The second independent variable Pro Change captures an actor’s position on an issue, which can be in favour or against policy change. This binary measure is based on a self-reported statement in a written submission to the hearing. Krippendorff’s alpha reports an acceptable score of .86 for this variable. Lastly, Public Support is measured as the proportion of the public in the opinion poll on the issue that shares the same position as an actor on the issue (based on the coded position in the hearing).
3.3.5 Control Variables
100
random intercepts for policy issues to account for the heterogeneity of different issues and for over-dispersion of the count measure.22
3.4 Analysis
Table 3.1 provides descriptive statistics on information provision by different types of actors. As the right column shows, information on public preferences is used by 45% of the actors. The figures in the left columns indicate that citizen groups are more likely to inform policymakers about public preferences: While 64% provide this information, approximately 44 % of professional groups and business groups and 30% of experts supply this type of information. Ultimately, however, the figures show that groups do transmit information about public preferences and have hence the potential to act as transmission belts.
Table 3.1: Provision of information on public preferences for different types of advocates (in percentages)
Citizen Groups Professional Groups Business Groups Experts & Others Total Informing about public preferences 64.29 43.90 44.00 30.00 45.22 Total N 84 82 100 90 356
Table 3.2 presents the findings to test hypotheses 1-3. As predicted in hypothesis 1, the negative coefficients in Models 1 and 2 (adding control variables) indicate that professional groups, business groups and experts provide significantly less information on public preferences than citizen groups. Model 2 shows that the differences for professional groups and business groups are significant at p<0.05 and for experts at p<0.001. Marginal predicted mean counts for different types of actors (based on Model 2) reveal that on average citizen groups provide information on public preferences 1.3 times per statement, while the amount for the other types of actors is between 0.66 and 0.86.
22 Alternatively, a zero inflated model could be used as the number of actors not providing political information
101
Table 3.2: Multilevel Negative Binomial Regression with random intercepts for policy issues (SEs in parentheses)
Model 1 2
Dependent Variable Info on Public Preferences Info on Public Preferences
H1: Actor Type (Ref: Citizen Groups)
Professional Groups -0.51** -0.44*
(0.19) (0.18)
Business Groups -0.49** -0.44*
(0.19) (0.18)
Experts and others -0.77*** -0.71***
(0.21) (0.21) H2: Pro Change -0.60** -0.60*** (0.18) (0.18) H3: Public Support 0.96** 1.04** (0.37) (0.35) Control Variables Number of Arguments 0.14*** 0.13*** (0.01) (0.01) Technical information -0.19 (0.16) Policy Type (Ref: Redistributive) Distributive 0.13 (0.54) Regulatory -1.21* (0.49)
Media Saliency (logged) 0.10
(0.07)
Constant -1.09*** -0.65+
(0.29) (0.37)
lnalpha -2.01*** -2.04***
(0.59) (0.60)
Policy Issue Intercept Yes Yes
N Cases (Issues) 356 (34) 356 (34)
AIC 756 751
+ p<0.10, * p<0.05, ** p<0.01, *** p<0.001
102
A further exploration of the control variable technical information supports this finding. As Model 2 shows, actors that provide technical information are less likely to provide information on public preferences. When not controlling for number of arguments the effect of technical information is positive and significant at p<0.05 (not shown). This suggest that, generally, the higher the likelihood that an actor provides technical information the higher the likelihood that the actor provides information on public preferences. Yet, when considering the length of the contribution, the results indicate that the emphasis is really one-sided: The more technical information provided, the less information on public preferences is provided. Hence, when trying to explain information provision, it is crucial to also look at the other types of information that are provided as this ultimately affects the provision of a specific type. This could also explain why some other work has not found differences across group types as the relational aspect has not been considered or is difficult to capture with self-reported information provision, whereby actors can make less accurate estimations of how much a certain type of information was used or was considered important (De Bruycker 2016). It does not mean that business and professional groups do not provide information about public preferences, nor that their informational value is less, solely that citizen groups emphasise it more, possibly because it is their stock in trade as they are a legitimate source for such information.
one-103
sided, with predominantly opponents of policy change mobilising at higher rates in order to protect existing legislations (Baumgartner et al. 2009). Appendix E therefore provides an analysis controlling for the level of conflict amongst advocates on an issue. While the results show that actors provide less information about public preferences when they face less conflict, the control does not alter the results.
Lastly, it was argued that the amount of public support an actor enjoys affects information provision. In line with this hypothesis, Model 2 shows a positive and significant relationship between public support and the amount of information the actor provides (p<0.01). Thus, the higher the share amongst the public having the same view as an actor, the more information on public preferences are transmitted by that actor. Figure 3.1 shows the predicted mean counts of information on public preferences for different levels of public support with 95% confidence intervals (based on Model 2).
Figure 3.1: Predicted Counts for Public Support for an Actor
104
public support actors have, the more they actually push for it. This may underline the strategic usage of this type of information and adds to studies that have shown that interest groups are more successful when they have the public on their side (Rasmussen, Mäder, et al. 2018). All findings are robust to controlling for a number of factors. As expected, actors provide more information about public preferences on redistributive issues compared to regulatory ones (p<0.05), possibly because the conflictual nature of such issues incentivises advocates to transmit their constituents’ interest. Furthermore, there is a highly significant effect (p<0.001) for the overall number of arguments made by an actor, i.e., the more arguments an actor makes the more information on public preferences is provided.
3.5 Alternative Model Specifications and Limitations
As mentioned, Appendix D provides an analysis using a binary measure indicating whether an actor transmitted public preferences or not. While the results for public support are not significant, all other findings show the same results. It suggests that public support is more important for the amount of information and less for whether to provide the information at all. This also fits to the caveat mentioned earlier: Minority groups are not less likely to provide information about their constituents when they have no public support, yet advocates provide it more frequently, the more their claim is supported by the general public.
Furthermore, the argument has been that interest groups act as transmission belts by informing both about general public opinion as well as specific constituents’ interest. Interestingly, advocates primarily make references to specific constituents and not public opinion at large. This shows that interest groups use public hearings to provide quite specific, probably privately held, information about their constituents that policymaker cannot easily access by other means such as the media or party colleagues. Appendix G therefore presents the main analysis using one proxy only and shows that the results are even stronger when looking at information on preferences of specific segments of society only.
105
issues, which often stimulate larger public interest than other types of issues. Taken together, this could mean that information about public preferences was easier to access for advocates on the policy issues in the sample. Furthermore, it could imply that the level of conflict on these issues is higher than average due to higher mobilisation. Yet, it also suggests that on issues the public cares about, interest groups take on their concerns and transmit their preferences.
Second, while the measure of information about public preferences allowed for gauging both references to public and constituency specific opinion, it only considers such references if the actor specifically referred to the public or a specific group of people. Obviously, groups can represent the public’s interest also by providing technical information. Even though this would not be counted as representation in Saward’s sense, it does not mean that such actors do not act in the interest of a constituency. This could imply that also business groups and professional groups transmit more constituency preferences than this study might lead us to expect. Yet, the same could be said for citizen groups, that is, the measure used in this study might also miss more of their attempts to act as representatives by providing technical expertise. In fact, we do not see significant differences amongst citizen groups and business and professional groups when looking at the amount of technical information (not shown) they provide, which supports existing research (De Bruycker 2016; Nownes and Newmark 2016) and suggests that citizen groups make use of it to a similar extent.
3.6 Conclusion
106
preferences in public hearings. In addition to expanding on the limited body of knowledge on information transmission of public preferences, the study’s content coding of observed information provides a more detailed measure of information transmission than existing studies that rely on surveys and interviews.
The results show that citizen groups transmit more information on public preferences than professional groups, business groups and experts. Thus, those that are seen as important surrogates of the public do transmit these preferences and have the potential to act as information providers that help public preferences get transmitted to policymakers. A recent study shows that groups vary somewhat in the extent to which they share the same view as the majority of the public (Flöthe and Rasmussen 2019). It is the same type of actors that is more likely to share the same view as the public that is also more likely to transmit information on public preferences to the policymaking level, suggesting that citizen groups are better able to represent the public both in substantive terms as well as in the sense of representative claims. Yet also those for whom representation is a by-product and who are often accused of dominating the interest group landscape transmit preferences. Interestingly, professional groups do not differ from business groups in their provision of information on public preferences. Furthermore, opponents of policy change provide more information on public preferences than supporters. Hence, if interest groups perceive the public interest is at risk, they inform policymakers about these negative consequences, which could be a potential mechanism driving the status quo bias as policymakers may be especially keen to avoid such risks. Lastly, the study shows that the more people share the same view as an actor, the more information on these preferences is provided. This underlines the potential of groups to act according to the wishes of the public and pushing for these preferences.
107
3.7 Appendices
Appendix A: List of Policy Issues on which Public Hearings were held
Cutting social benefits
Cutting budget for infrastructure Cutting budget for Children Support
Implementing financial transaction tax at national level
Punishment for people evading taxes even though they self-reported Regulating salaries of managers
Privatization of Deutsche Bahn
Decreasing Taxes to boost the economy Financial release for companies
Raising unemployment benefits according to the development of living costs Minimum wage for delivery mail service sector
Increasing fees for health insurance Minimum wage for all sectors Establishing a wealth tax Introducing a unique tax rate Cutting coals subsidies Increasing Taxes
Making it illegal to carry out a paternity test without the consent of the mother Introducing a unique health insurance (Citizen Insurance)
Prosecuting black labour in private households
Raising the fee for care insurance for people without children Abolishing housing grants
Reducing commuter tax
Health insurance should only be paid by employees and not together with employers Making a dentures insurance obligatory
Introducing a fee for patients when visiting the doctor Citizen should pay 10% of their health service Abolishing tax breaks
Raising tobacco tax
Cutting unemployment benefits No gradual increase for pensions Raising pension contributions
Increasing the retirement age up to 67
Facilitate moving to Germany for foreign employees
108
Appendix B
B 1: Variable Overview
Variable Obs Mean Std. Dev. Min Max
Information on Public Preferences (Count)
356 8792135 1.502867 0 11
Pro change (binary) 356 0 1
Group Type: Citizen Groups Professional Groups Business Groups Experts 356 1 4
Public Support for Actor 356 .5584951 .229845 .0703278 .9456077
Number of Arguments 356 4.657303 4.689272 0 28
Provision of Technical Information 356 0 1
Policy type: Distributive Redistributive Regulatory
356 1 3
Media Saliency (logged)
(average number of articles per day)
109
Appendix C
C1: Coding Instructions for Public Hearings23
Identifying an argument
In order to identify an argument, the coder should look for an actor’s opinion on an issue, likely to be signalled by “demands, criticisms” related to the issue (Koopmans and Statham 1999). In order to identify and code an argument, it helps to think what the causal story is behind one’s statement, i.e., actors claiming a position and giving a reason for their position (“we support this, because of that”). Second, after identifying the position on the issue, it is crucial to identify the different types of arguments used. The coder should read relevant passages after and before the statement. Different kinds of arguments can belong to the same type of information. “The proposal will lead to a loss of 300.000 jobs” and “the proposed policy will result in an increase of our greenhouse gas emissions of 20%” are two different arguments, both belonging to the broader category of “information regarding impact for economy, environment and alike”. Any change of the structure of the arguments and/or the semantic meaning signifies a new argument and should be counted separately. Often actors also use signal words such as “additionally”, “furthermore” etc. in order to add an additional argument. Often, a new argument is presented in a new paragraph. If the same argument re-occurs at another place, it has to be coded and counted again. Below is a figure visualising the relationship between the concepts.
23The coders were instructed to code 13 different arguments in order to capture different types of information.
110 Coding of Arguments
111
Political Expertise
1.1 References to public opinion
Does the actor make any references how much public support the proposal would have? Does the actor signals support or opposition from subparts of the public?
Examples: “This proposal is highly unpopular amongst the public.” “70% of the public support the idea.” “All sectors are against the proposal.”
1.2 References to whose interest would be affected
Does the actor argue how the proposed policy will affect certain subgroups of the society? Are there references to who would benefit and who would be harmed?
Example: „This measure leads to higher burdens on contributors and is therefore to be rejected. “ „It is important to point out that the situation for families with a low income would be aggravated.”
Policy Expertise
2.1 Facts and Technical Knowledge
Does the actor present studies and analysis (conducted by him or herself or the organisations he or she is working in or external references)? Does the actor present facts about the current situation and facts about the impact of the policy based on calculations? Does the actor present very detailed and highly technical information?
112
C2: Interest group type categorization 24
1 Labour groups 11 Blue-collar union 12 White-collar union
13 Other labour groups (i.e. think tanks related to unions) 14 Employee representative committee
2 Business groups
21 Peak-level business group 22 Sector-wide business group 23 Breed associations
24 Technical Associations 25 Other business group 3 Institutional Associations
31 Associations of local authorities
32 Associations of other public institutions
33 Associations of managers of public institutions 34 Other Institutional associations
Occupational associations 41 Doctors’ associations 42 Other medical professions 43 Teachers’ associations
44 Other occupational associations Identity Groups
51 Patients 52 Elderly 53 Students
54 Friendship groups (i.e. non-specific groups related to a country) 55 Racial or ethnic
56 Other – undefined - identity group 57 Women 58 Lesbian/Gay/Bisexual/Transsexual Hobby/Leisure groups 61 Sport 62 Other hobby/leisure Religious groups
24 In the case of labour groups, business groups, and institutional associations, the subgroup does not need to be
113
71 Associated with the protestant church 73 Other religious group
74 Roman/Catholic groups Public interest groups
81 Environment and animal welfare 82 Humanitarian – international 83 Humanitarian – national 84 Consumer Group
85 Other – undefined – public interest 86 Government reform
87 Civil liberties
88 Citizen Empowerment 90 Think Tanks
114
Appendix D-G: Different Model Specifications
Table D: Multilevel Logit Models with random intercepts for policy issues and use of the binary Dependent Variable (SEs in Parentheses)
Model D1 D2
Dependent Variable Information on Public
Preferences
Information on Public Preferences H1: Actor Type
(Ref: Citizen Groups)
Professional Groups -0.88* -0.77*
(0.38) (0.37)
Business Groups -1.25** -1.09**
(0.39) (0.39)
Experts and others -1.39*** -1.28***
(0.40) (0.39) H2: Pro Change -1.01** -0.97** (0.31) (0.30) H3: Public Support 0.65 0.98 (0.69) (0.69) Control Variables Number of Arguments 0.26*** 0.24*** (0.05) (0.05) Technical information -0.36 (0.36) Policy Type (Ref: Redistributive Distributive -0.38 (0.96) Regulatory -1.62** (0.63)
Media Saliency (logged) 0.11
(0.14)
Constant -0.45 0.02
(0.56) (0.79)
Policy Issue Intercept Yes Yes
N Cases (Issues) 356 (34) 356 (34)
AIC 393 392
+ p<0.10, * p<0.05, ** p<0.01, *** p<0.001
115
Appendix E: Controlling for Conflict amongst Advocates
116
Table E: Multilevel Negative Binomial Regression with random intercepts for policy issues (SEs in parentheses) predicting Information on Public Preferences (count). Controlling for Conflict amongst actors on an issue.
(E1)
Information on Public Preferences H1: Actor Type
(Ref: Citizen Groups)
Professional Groups -0.43*
(0.18)
Business Groups -0.45*
(0.18)
Experts and others -0.72***
(0.21) H2: Pro Change -0.73*** (0.19) H3: Public Support 1.31*** (0.35) Control Variables Number of Arguments 0.13*** (0.01) Technical information -0.21 (0.15)
Conflict amongst advocates -1.04*
(0.46) Policy Type (Ref: Redistributive)
Distributive 0.07
(0.53)
Regulatory -1.21*
(0.48)
Media Saliency (logged) 0.11
(0.07)
Constant -0.48
(0.36)
lnalpha -2.17***
(0.63)
Policy Issue Intercept Yes
Number of Cases 356
AIC 746
117
Appendix F: Interaction between Group Type and Public Support
118
Table F: Multilevel Negative Binomial Regression with random intercepts for policy issues (SEs errors in parentheses) predicting Information on Public Preferences (count) and interacting Public Support and Group Type
Model F1 F2
Dependent Variable Info on Public Preferences Info on Public Preferences
Actor Type
(Ref: Citizen Groups)
Professional Groups -0.28 -0.02
(0.76) (0.76)
Business Groups -0.33 -0.29
(0.60) (0.60)
Experts and others -0.02 0.16
(0.68) (0.68)
Pro Change -0.57**
(0.18)
Public Support 1.40+ 1.49*
(0.75) (0.74)
Group Type* Public Support Professional Groups* Public Support -0.44 (1.08) -0.60 (1.06) Business Groups* Public Support -0.12 (0.90) -0.14 (0.89) Experts*Public Support -1.56 -1.50 (1.07) (1.06) Control Variables Number of Arguments 0.15*** 0.13*** (0.01) (0.01) Technical information -0.19 (0.16) Policy Type (Ref: Redistributive) Distributive 0.14 (0.54) Regulatory -1.22* (0.48)
Media Saliency (logged) 0.10
(0.07)
Constant -1.56** -0.97+
(0.53) (0.58)
lnalpha -1.89*** -2.00***
(0.55) (0.59)
Policy Issue Intercept Yes Yes
N Cases (Issues) 356 (34) 356 (34)
119
Figure F2: Predicted Counts for Public Support by Actor Type
120
Table G: Multilevel Negative Binomial Regression Analysis with random intercepts for policy issues predicting Information on Public Preferences using only one proxy
(references to specific segments of society) (SEs in Parentheses)
Model G1 G2
Dependent Variable Info on Public Preferences Info on Public Preferences
H1: Actor Type (Ref: Citizen Groups)
Professional Groups -0.57** -0.49**
(0.19) (0.19)
Business Groups -0.51** -0.46*
(0.19) (0.19)
Experts and others -0.95*** -0.86***
(0.23) (0.23) H2: Pro Change -0.80*** -0.80*** (0.20) (0.20) H3: Public Support 1.25** 1.28*** (0.38) (0.36) Control Variables Number of Arguments 0.14*** 0.13*** (0.01) (0.01) Technical information -0.16 (0.16) Policy Type (Ref: Redistributive) Distributive -0.45 (0.73) Regulatory -1.56** (0.60)
Media Saliency (logged) 0.10
(0.07)
Constant -1.31*** -0.83*
(0.30) (0.37)
lnalpha -2.04** -2.13***
(0.68) (0.64)
Policy Issue Intercept Yes Yes
N Cases (Issues) 356 356
AIC 700 689
